The present invention relates to a system utilization rate managing apparatus and a system utilization rate managing method to be employed for it, and its program, and in particular, to a managing method of managing a performance and a utilization rate of a node group for a system that is composed of a plurality of calculating nodes.
Conventionally, the system utilization rate managing apparatus includes a processing device 51 and a memory device 52, the processing device 51 includes a CPU (Central Processing Unit) utilization rate acquirer 511, a CPU utilization rate comparator 512, and a HW (Hardware) configuration alteration determinator 513, and the memory device 52 includes a CPU utilization rate memory section 521 and a service level contract memory section 522, respectively, as shown in FIG. 9 (For example, refer to Patent Document 1).
The conventional system utilization rate managing apparatus 5 having the configuration as mentioned above operates as follows. First, the CPU utilization rate acquirer 511 acquires a CPU utilization rate of each node from an external system (not shown) composed of one node or more that is not shown in the figure, and files it into the CPU utilization rate memory section 521. An upper limit of the CPU utilization rate that each node should maintain, a maximum number of the nodes that are reinforceable in cases where the CPU utilization rate of a node has exceeded this upper limit, etc. are prestored in the service level contract memory section 522.
The CPU utilization rate comparator 512 compares the CPU utilization rate of each node, which the CPU utilization rate memory section 521 retains, with the upper limit of the CPU utilization rate that should be maintained, which the service level contract memory section 522 retains, and gives a comparison result to the HW configuration alteration determinator 513. If there is a node of which utilization rate exceeds the upper limit, the HW configuration alteration determinator 513 determines that the service level is not satisfactory, and reinforces the node for the external system. Nodes can be reinforced up to the maximum number of the reinforceable nodes that the service level contract memory section 522 retains.
In contrast if there is a node of which utilization rate is equal to or less than the upper limit of CPU utilization rate, the HW configuration alteration determinator 513 determines that the service level of the external system is satisfactory. If the service level can be maintained with less nodes, the node may be curtailed. The HW configuration alteration determinator 513 estimates the utilization rates after reinforcement or curtailment by regarding the product of the utilization rate and the number of nodes as constant under the assumption that CPU power of all nodes are identical. For example, when four nodes are all operated at the CPU utilization rate of 25%, the utilization rate of each node at the time of having increased the number of the node to five is estimated to be 20%. Conversely, the utilization rate of each node at the time of having curtailed the number of the node to two is estimated to be 50%.
[Patent Document 1] JP-P2002-24192A
With the foregoing system utilization rate managing apparatus, there is the problem that it cannot be properly estimated whether nodes should be reinforced in cases where current nodes have different performances. The reason is that the foregoing apparatus cannot estimate the total performance or the total capacity of nodes of various performances. For example, if the utilization rate of a node of low performance is higher than an upper limit, and the utilization rate of another node of high performance is quite low, then the system including the two nodes has large capacity. Nevertheless, the foregoing apparatus will reinforce nodes to the system in spite of the large capacity.
Further, in many cases, CPU clock, which are simple metrics of performance, does not accurately signify the performance of an application software operating over the system. This is because the performance of an application does not always depends on CPU clocks but on the architecture of CPU and the entire computer system.